@mux/ai

Per-title encoding is

a technology

and a video

encoding approach

originally proposed

by Netflix in 2015.

The fundamentals

of this technology

is that no

two videos are the same.

So you shouldn't

encode them the same way.

For example,

this video of a waterfall

is really different to

this video

of a tiny hippopotamus

and so

you should encode

those videos differently.

Originally, Netflix proposed

running a

whole bunch of test encodes

on every video file.

Then, measuring the quality

of all of those encodes

to decide what the best

encoding settings

were for that content.

But that's pretty wasteful.

Over the last

nearly ten years

now, people have found

all sorts

of different approaches

to build per-title

encoding algorithms.

For example,

You could use an AI algorithm

that classify content

into different categories

like “sports”

or “talking head” video,

and then pick

an appropriate ladder

for that content.

Mux’s approach to per-title

encoding is different.

We actually look

at the complexity information

encoded within the video.

So that waterfall video

we looked at earlier

has a lot of complexity.

It's got a lot of motion,

in a lot of detail.

Whereas that hippopotamus,

not so much.

We take that

complexity information

and run it through a magical

AI algorithm,

which calculates all the best

encoding settings

for your video,

including CRF

values, bitrate values,

decoder buffers, you name it,

we calculate it.

Put all of that

together and your video,

no matter how complex it is,

can always look

great on Mux.